I have a problem understanding how a DMZ would be implemented. I've read two different answers regarding the communication between the DMZ and the internal network. Some people suggest that you can open some connections from the DMZ to the some services on the internal network. Other people suggest that absolutely no communication should be allowed from the DMZ to the LAN.
Suppose that I want to implement a webserver on the DMZ that communicates with a database thats inside the LAN. How would i go about this? How would the webserver read the necessary info from the DB without allowing connections from the DMZ to the DB, at least on that single DB port? I've read some people suggesting periodic updates from the DB to the webserver, but that seems like a terribly inefficient solution.